Self-Driving Money

The idea, and why it stalled

About a decade ago, fintech had a shared notion, in Eric Glyman's own words: "one day there will be self-driving money, companies and people will have their assets in the highest-yielding security or asset at all times, not waste a dollar on things they don't need, and have better insights on how to live a better financial life."¹ He is blunt about the outcome: it "didn't happen." The missing pieces were the same ones that blocked self-driving cars for years, the world was not yet digitized or connected enough, and there was no single command-and-control layer with enough context to act safely.¹

Why it becomes real now

Glyman's claim is that the preconditions have finally arrived. Businesses are increasingly digitized, so transactions, receipts, books, and HR data are machine-readable, making it easier to know with accuracy what actually happened. They are increasingly connected, with a platform like Ramp stitching card, bill pay, reimbursements, treasury, and HR data into a single model. And a command-and-control system now exists that can sit across those money flows and both decide and act, not merely observe.¹

Given those three conditions, "self-driving money is actually kind of a real thing, in the same way a decade ago people were excited about self-driving cars, and it's finally here and people are just getting used to it." Glyman frames the outcome as inevitable regardless of who builds it: "whether it's us or someone, it's going to happen, and we want to be the company that does that."¹

Buying work, not seats

Glyman has also put the underlying business logic plainly: "you're not buying seats, software, you're buying work."² Two details sharpen the vision. The first is a Tesla analogy: a Tesla is designed not just to get a driver from one place to another but to drive them there without requiring their focus, and the same logic applies to financial processes, tapping a card while the system runs a real-time policy check, pulls merchant data, writes the memo, and posts to the ledger automatically. The second is a direct statement of the end state: "It's crazy, all the data is digital in the first place... you're an adult allowed to buy things on behalf of your company; expenses should do themselves. If computers can think, your books should close themselves."² The resulting shift in human work moves finance professionals from mostly backward-looking recurring tasks toward deciding who the customer is and where the next dollar should go.

Why it matters

Self-driving money is the outcome layer of a fintech company's strategy made concrete: not a tool a person operates, but a system that performs the financial work itself. It is the same underlying arc as the self-driving-car analogy used elsewhere in fintech: agentic products that quietly issue and configure their own cards, move idle cash to the highest available yield, and reduce the amount of direct human operation required to run a company's finances.

Tensions

Trust and liability remain the open constraint: self-driving finance means an agent moving real money under real regulatory exposure, and the current compromise is closer to a supervised alpha product than a fully autonomous end state. And the promise of being in the highest-yielding place at all times assumes that risk tolerance, liquidity needs, and covenants can all be reliably encoded, which is the same edge-case depth that makes any fully autonomous treasury system difficult to build in practice.